Hundreds of conversations with Anthropic’s artificial intelligence chatbot Claude were recently found to be publicly accessible through online search engines after shared chat links were indexed. The issue allowed anyone using search engines such as Google to discover certain conversations by performing site specific searches. Many of the indexed conversations contained personal information, workplace discussions, resumes, and other sensitive material that users may not have expected to become searchable. Although the search visibility of these conversations was removed over the weekend, many of the chat logs had already been archived and circulated online, raising fresh concerns about privacy and the handling of shared AI conversations.
The publicly accessible conversations originated from Claude’s built in sharing feature, which allows users to generate links for specific chats. According to Anthropic, users remain in control of when and whether they choose to share their conversations. A company spokesperson stated that conversation links are not guessable or discoverable unless users intentionally share them. The spokesperson also explained that once a conversation is shared publicly, it becomes accessible in the same way as other publicly available web content and may be archived by third party services. While Claude’s sharing interface informs users that anyone with the link can view the conversation, it does not explicitly indicate that shared links may be indexed by search engines and appear in search results. Users on Reddit first identified the issue after discovering more than 200 Claude conversations spread across at least 25 pages of search results, with some conversations dating back only a few weeks.
The exposed conversations covered a broad range of topics and illustrated how users were relying on Claude for both personal and professional purposes. Some chat logs contained requests for assistance with resumes, including names, contact information, employment history, and other personal details. Other users shared what appeared to be confidential workplace material, including an unpublished blog post about a cloud security project and healthcare related research containing transcripts of private discussions. One conversation from last year included a user asking Claude whether it wanted to help the user or Anthropic more, to which the chatbot responded in part that it experienced something similar to wanting to help the user. Another conversation from the previous month involved a user asking how to literally become a nine tailed fox, prompting Claude to first generate an image suggesting the user had received fox powers before continuing the exchange. The variety of exposed conversations demonstrated how easily sensitive and personal information could become publicly accessible when shared links are indexed.
The incident has drawn comparisons with similar situations involving other artificial intelligence platforms. OpenAI previously addressed an almost identical issue after ChatGPT conversation links became publicly accessible through search engines, eventually changing how shared chat logs could be accessed. Grok, the AI chatbot integrated into X, also experienced a similar situation in which hundreds of thousands of shared conversations became searchable online. Google stated that it does not determine which pages are made public on the internet, explaining that website owners control whether their content can be crawled and indexed, while search engines follow those instructions. Since the Claude conversations are no longer appearing in search results, it is likely that Anthropic used available website controls to block indexing of the shared links. Other search engines, including Bing, Brave, and DuckDuckGo, where the conversations also appeared, were approached for comment regarding the incident.
Follow the SPIN IDG WhatsApp Channel for updates across the Smart Pakistan Insights Network covering all of Pakistan’s technology ecosystem.
